Install speech with Homebrew

Project links

• Homepage
• Repository

Commands and aliases

• speech
• speech-server

Project history

The official repository was created in February 2026 and describes Speech Swift as AI speech models for Apple Silicon, powered by MLX Swift and CoreML, for Mac and iOS speech recognition, synthesis, and understanding. The README emphasizes local execution with no cloud service, API keys, or off-device data transfer.

Adoption history

The project moved quickly through 0.0.x releases in 2026. Its v0.0.20 release notes say the CLI binary was renamed from audio to speech specifically to make room for an eventual Homebrew formula, which ties the package name directly to command-line distribution concerns.

How it is used

The README presents the toolkit as both Swift package modules and command-line demos; release notes show Homebrew-style usage such as upgrading the speech formula or building the release binary from source. Package users care because it turns large on-device speech models into a normal installable CLI on Apple Silicon.

Why package nerds care

For package maintainers, speech is interesting because it represents the 2020s shift from cloud speech APIs to local ML runtimes installed with ordinary package managers. Its formula name, binary rename, model assets, and Apple Silicon focus are all practical packaging concerns rather than just library metadata.

Timeline

• 2026: soniqo/speech-swift repository created on GitHub.
• 2026: v0.0.20 renamed the CLI binary from audio to speech for Homebrew packaging.
• 2026: v0.0.21 release notes continued documenting CLI and Homebrew upgrade usage.

Related projects

• MLX Swift and CoreML provide the local Apple Silicon execution backends used by Speech Swift.
• Speech Swift wraps several speech model families, including ASR, TTS, VAD, and diarization components.

Sources

• https://api.github.com/repos/soniqo/speech-swift
• https://github.com/soniqo/speech-swift
• https://github.com/soniqo/speech-swift/releases